- Feb 1, 2017
There is no info on which kind of values NN layer of TAA upscaling used in DLSS 2.0 is using.Tensor cores performing int8 inference
GPUs are inherently well suited for NN inference, as NV itself was shouting loudly in the past, way before TCs.
TCs are needed to explain why something is disabled on older cards and "I just want you to buy my newer product" wouldn't fly.
FidelityFX is cross platform.I'm just warning everyone who's worshipping at the alter of Lisa Su that they don't care (much) more about you than Nvidia does
Freesync is cross platform.
Mantle turned into Vulkan is... you've guessed it.
Meanwhile Huang managed to kill OpenGL by pushing proprietary crap. (Microsoft was amused, I guess)
So, Lisa over Jensen any time, thank you very much.
This "but they are commercial companies, so they would do anything to make more money" is getting old.
Yes, they are commercial companies, and one of them is used to all kind of shit, while another plays classy.
If Huang tells you something is two times faster, it is highly unlikely it is.
Whereas when Dr Su tells you, something is X% faster, you could be sure it really is.